aymenfurter/polyclaw/foundry-agent-chat

Name: foundry-agent-chat
Author: aymenfurter

Claude Code and agent skill by aymenfurter/polyclaw

Trust tier 2

community

LLM review pending

By aymenfurter
Coding
aymenfurter/polyclaw
Plugin foundry-agents
plugins/foundry-agents/skills/foundry-agent-chat/SKILL.md
Updated 25 Mar 2026
Version 1.0.0

Source check pending

Checking that the SKILL.md file is still in the repository.

Create an ad-hoc Microsoft Foundry agent for a specific purpose using the new v2 Responses API. Build a custom prompt agent with optional tools, run a conversation, and clean up. Use when the user says: "create a foundry agent", "make an agent for", "foundry agent", "build an ai agent", "create an assistant", "spin up an agent", "agent for this task"

Install

npx vskill@latest install aymenfurter/polyclaw/foundry-agent-chat

SKILL.md

Foundry Agent Chat

Create an ad-hoc Foundry prompt agent tailored to a specific purpose using the new v2 Responses API with conversations.

Prerequisites

The user must have completed the Foundry setup (the setup-foundry skill). They need:

FOUNDRY_PROJECT_ENDPOINT -- the project endpoint URL
FOUNDRY_MODEL_DEPLOYMENT_NAME -- the model deployment name

If these are not set, ask the user for the values.

Steps

1. Define the Agent Purpose

Ask the user:

What should the agent do? (e.g., "Write marketing copy", "Answer questions about our docs", "Translate text")
Should it have any special instructions or persona?
Does it need any tools? (code interpreter, file search, web search)

2. Create and Run the Agent

Use the following Python script. Customize the agent_name, instructions, and tools based on the user's requirements.

Basic prompt agent (no tools):

bash

python3 << 'PYEOF'
import os
from azure.identity import DefaultAzureCredential
from azure.ai.projects import AIProjectClient
from azure.ai.projects.models import PromptAgentDefinition
 
endpoint = os.environ.get("FOUNDRY_PROJECT_ENDPOINT", "<ENDPOINT>")
model = os.environ.get("FOUNDRY_MODEL_DEPLOYMENT_NAME", "<MODEL>")
 
project_client = AIProjectClient(
    endpoint=endpoint,
    credential=DefaultAzureCredential(),
)
 
with project_client:
    openai_client = project_client.get_openai_client()
 
    # Create a purpose-built agent
    agent = project_client.agents.create_version(
        agent_name="AGENT_NAME_HERE",
        definition=PromptAgentDefinition(
            model=model,
            instructions="INSTRUCTIONS_HERE",
        ),
        description="DESCRIPTION_HERE",
    )
    print(f"Agent created: {agent.name} v{agent.version}")
 
    # Create a conversation
    conversation = openai_client.conversations.create()
 
    # Send the user's message
    response = openai_client.responses.create(
        conversation=conversation.id,
        input="USER_MESSAGE_HERE",
        extra_body={"agent": {"name": agent.name, "type": "agent_reference"}},
    )
 
    # Print the response
    print(response.output_text)
 
    # Clean up
    project_client.agents.delete_version(agent_name=agent.name, agent_version=agent.version)
    print("Agent cleaned up.")
PYEOF

Agent with tools (code interpreter + web search):

bash

python3 << 'PYEOF'
import os
from azure.identity import DefaultAzureCredential
from azure.ai.projects import AIProjectClient
from azure.ai.projects.models import (
    PromptAgentDefinition,
    CodeInterpreterTool,
    CodeInterpreterToolAuto,
)
 
endpoint = os.environ.get("FOUNDRY_PROJECT_ENDPOINT", "<ENDPOINT>")
model = os.environ.get("FOUNDRY_MODEL_DEPLOYMENT_NAME", "<MODEL>")
 
project_client = AIProjectClient(
    endpoint=endpoint,
    credential=DefaultAzureCredential(),
)
 
with project_client:
    openai_client = project_client.get_openai_client()
 
    # Create agent with tools
    agent = project_client.agents.create_version(
        agent_name="AGENT_NAME_HERE",
        definition=PromptAgentDefinition(
            model=model,
            instructions="INSTRUCTIONS_HERE",
            tools=[
                CodeInterpreterTool(container=CodeInterpreterToolAuto()),
                {"type": "web_search_preview"},
            ],
        ),
        description="DESCRIPTION_HERE",
    )
    print(f"Agent created: {agent.name} v{agent.version}")
 
    # Create a conversation
    conversation = openai_client.conversations.create()
 
    # Send the user's message
    response = openai_client.responses.create(
        conversation=conversation.id,
        input="USER_MESSAGE_HERE",
        extra_body={"agent": {"name": agent.name, "type": "agent_reference"}},
    )
 
    # Print the response
    for item in response.output:
        if item.type == "message":
            for content in item.content:
                if content.type == "output_text":
                    print(content.text)
 
    # Clean up
    project_client.agents.delete_version(agent_name=agent.name, agent_version=agent.version)
    print("Agent cleaned up.")
PYEOF

3. Multi-turn Conversation

If the user wants to continue the conversation, reuse the same conversation ID without creating a new one:

bash

python3 << 'PYEOF'
import os
from azure.identity import DefaultAzureCredential
from azure.ai.projects import AIProjectClient
 
endpoint = os.environ.get("FOUNDRY_PROJECT_ENDPOINT", "<ENDPOINT>")
 
project_client = AIProjectClient(
    endpoint=endpoint,
    credential=DefaultAzureCredential(),
)
 
with project_client:
    openai_client = project_client.get_openai_client()
 
    # Reuse existing conversation and agent
    conversation_id = "CONVERSATION_ID_HERE"
    agent_name = "AGENT_NAME_HERE"
 
    response = openai_client.responses.create(
        conversation=conversation_id,
        input="FOLLOW_UP_MESSAGE_HERE",
        extra_body={"agent": {"name": agent_name, "type": "agent_reference"}},
    )
 
    print(response.output_text)
PYEOF

4. Present Results

Show the agent's response to the user
If the agent used tools (code interpreter, web search), summarize what it did
For multi-turn conversations, keep track of the conversation ID and agent name

5. Clean Up

When the user is done, delete the agent version:

bash

python3 << 'PYEOF'
import os
from azure.identity import DefaultAzureCredential
from azure.ai.projects import AIProjectClient
 
endpoint = os.environ.get("FOUNDRY_PROJECT_ENDPOINT", "<ENDPOINT>")
 
project_client = AIProjectClient(
    endpoint=endpoint,
    credential=DefaultAzureCredential(),
)
 
with project_client:
    project_client.agents.delete_version(agent_name="AGENT_NAME", agent_version="VERSION")
    print("Agent cleaned up.")
PYEOF

Agent Types Summary

Use Case	Tools	Example
Q&A / Chat	None	Customer support bot, writing assistant
Data Analysis	Code Interpreter	CSV analysis, chart generation, math solver
Research	Web Search	Market research, fact checking
Full-featured	Code Interpreter + Web Search	Research analyst with computation

Notes

Agents are versioned: each create_version call increments the version number
Conversations persist state across calls, so the agent remembers context
Use store=False in the response call to opt out of statefulness when not needed
The Responses API is synchronous -- no polling required (unlike the old Runs API)
Agent definitions and conversation state use single-tenant storage for security

Security

Security scanPassed, 100 of 100Pattern checks on the skill's files.
Model reviewNot runA model review is scheduled for the next rescan.

See the full security report

Quality evaluation

Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

See evaluation results

Source

SKILL.md in the repositoryCheckingplugins/foundry-agents/skills/foundry-agent-chat/SKILL.md

Popularity

Repository stars: 50
Repository forks: 8
7-day trend score: 20

Stars and forks are for the whole GitHub repository, not this skill alone. Refreshed 20 Mar 2026, 7 months ago.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

Claude Code
Cursor
GitHub Copilot
Windsurf
Codex
Gemini CLI
Cline
Amp
Roo Code
Goose
Continue
Kilo Code
Junie
Trae
OpenHands
Qwen Code
AI Agent
Tabnine Chat
Sourcegraph Cody
Amazon Q
Firebase Genkit
Sweep
Devin
Phind
Mentat
Claude Desktop
Aider
GPT Pilot
Plandex
bolt.new
v0
Replit Agent
Anse
TypingMind
BetterChatGPT
LibreChat
Warp AI
Pieces
Fig AI

Embed badge

[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/aymenfurter/polyclaw/foundry-agent-chat/badge)](https://verified-skill.com/skills/aymenfurter/polyclaw/foundry-agent-chat)

Something wrong with this skill or its page? Tell us.